Claude Privacy and Data Use

2:48 Claude uses your chats to train future models only if you allow it: on the consumer plans (Free, Pro, Max and Ultra) the choice is the model improvement toggle in Settings > Privacy, and on Team, Enterprise, Education and the Claude API Anthropic does not train on your content by default (Plans and pricing, read Nov 6, 2026). This has been the rule since Anthropic's consumer terms change of Aug 28, 2025. The trade-off is retention: with training on, Anthropic may keep chats for up to five years, and with it off, the standard 30 days. Incognito chats are never used for training.

Quick facts

Does Claude train on my chats?

It depends on your plan and one setting. Consumer accounts, including Claude Code used with those accounts, decide with the model improvement toggle. If it is on, your new and resumed chats and coding sessions may be used to train future Claude models; if it is off, they are not (Updates to Consumer Terms, Aug 28, 2025). Chats you had before choosing, and never reopened, are not used.

There are exceptions even with the setting off. A conversation flagged for a safety review may still be used to improve Anthropic's trust and safety systems, and rating a response (thumbs up or down) sends the whole conversation as feedback, which Anthropic keeps for five years (How do I change my model improvement privacy settings?; Privacy Policy, read Nov 6, 2026). Raw content from Connectors such as Google Drive is not used for training unless you paste it into the chat (Is my data used for model training?).

Anthropic's stated reason for the 2025 change was that models take 18 to 24 months to develop, so it wanted real conversations, and longer-held data, to train both models and its abuse classifiers (Anthropic). Existing users had to make a choice by Oct 8, 2025 to keep using Claude.

How to turn off training

1. In Claude on the web, desktop or mobile, open Settings > Privacy (on the web, claude.ai/settings/data-privacy-controls).
2. Turn off the model improvement toggle, labeled "Help Improve our AI models" in the Privacy Center's instructions.
3. The change applies to future training: Anthropic stops using your old and new chats for new training runs, but cannot remove them from runs already started or models already trained (Privacy Center, read Nov 6, 2026).

How long does Anthropic keep Claude chats?

For consumer accounts, the Privacy Center gives these periods (How long do you store my data?, read Nov 6, 2026):

Those Privacy Center articles still name only Free, Pro and Max, while newer Help Center articles also list Ultra as a consumer plan; the two had not been reconciled on Nov 6, 2026.

Team, Enterprise and the API

Anthropic's commercial products (Team, Enterprise, Claude Gov, Education and the API, including through Amazon Bedrock and Google Cloud) do not use inputs or outputs for training by default, unless the customer sends feedback or opts in (Is my data used for model training? (commercial)). Team and Enterprise owners can turn off thumbs-up and thumbs-down feedback for the whole organization under Organization settings > Data and Privacy. API inputs and outputs are deleted within 30 days, except for the Files API, zero data retention agreements, policy enforcement or legal holds; Enterprise can set custom retention (How long do you store my organization's data?).

0:19 One exception is new in 2026: Mythos-class models, such as Claude Fable 5, carry required limited retention and review as part of Anthropic's safety work, even for organizations (same).

Incognito chats

Incognito Chats are the private mode. They are not saved to your history or to Claude's Memory, are not used for training even with model improvement on, and are kept 30 days by default (Use incognito chats). Start one with the ghost icon at the top right of a new chat; it is not available inside a project. On Team and Enterprise plans, incognito chats still appear in the organization's exports and, on Enterprise, in the Compliance API, so they are private from your history, not from your employer.

Deleting and exporting your data

Deleting a chat removes it from your history at once and from Anthropic's back-end systems within 30 days; a deleted chat is not used for future training (Delete or rename a conversation). You can export everything from Settings > Privacy on the web or desktop app (not the phone apps), and delete your whole account from Settings > Account. Step-by-step instructions are on How to Export or Delete Claude Data.

Who is responsible for my data?

Outside Europe, the data controller is Anthropic PBC in San Francisco; for the European Economic Area, the UK and Switzerland it is Anthropic Ireland, Limited in Dublin (Privacy Policy, effective Sep 10, 2026). For how this compares with ChatGPT and Gemini, see AI Privacy Compared on ai.vi.ki, and for overall safety, Is Claude Safe.

Frequently asked questions

Does Claude train on my data? Only if you allow it on a consumer plan. Free, Pro, Max and Ultra users choose with the model improvement toggle in Settings > Privacy; Team, Enterprise and API data is not used to train Claude by default.

How do I stop Claude from using my chats for training? Open Settings > Privacy in Claude and turn off the model improvement toggle. Claude then stops using your chats for future training, though data already in a training run cannot be pulled out.

How long does Claude keep my conversations? Claude keeps chats until you delete them; deleted chats are purged within 30 days. With training on, de-identified copies may stay in training pipelines for up to five years.

Are incognito chats in Claude private? Claude's incognito chats are not saved to history or memory and are never used for training, but Anthropic keeps them for 30 days, and on Team and Enterprise plans they appear in the organization's exports.

Can my employer see my Claude chats? On Claude Team and Enterprise, owners can export members' chats, including incognito ones, and Enterprise admins can read them through the Compliance API; assume a work account is visible to your employer.

Does Anthropic sell Claude user data? No. Anthropic states that it does not sell Claude users' data to third parties.

Sources

• What does AI actually know about you?, Claude, Aug 13, 2026 (video)
• How data retention works when using Claude, Claude, Sep 14, 2026 (video)
• Updates to Consumer Terms and Privacy Policy, Anthropic, Aug 28, 2025
• Privacy Policy, Anthropic, effective Sep 10, 2026, read Nov 6, 2026
• Plans and pricing, Claude, read Nov 6, 2026
• How long do you store my data?, Anthropic Privacy Center, read Nov 6, 2026
• Is my data used for model training? (consumer), Anthropic Privacy Center, read Nov 6, 2026
• How do I change my model improvement privacy settings?, Anthropic Privacy Center, read Nov 6, 2026
• Is my data used for model training? (commercial), Anthropic Privacy Center, read Nov 6, 2026
• How long do you store my organization's data?, Anthropic Privacy Center, read Nov 6, 2026
• Use incognito chats, Claude Help Center, read Nov 6, 2026
• Delete or rename a conversation, Claude Help Center, read Nov 6, 2026